Re: [Gnumed-devel] Measurement conversion concepts

[Busser, Jim]

A few questions about lab test unit conversion (and normal range 
standardization) remain hanging.

I would appreciate if we can come back to them within the next couple of weeks:


1) re: notional (but as-yet unused) tables in the schema 'public':

        public.unit
        public.basic_unit

… do we wish to keep such tables in "public" or do they better fit in "ref"?

… maybe there is a tension here between "basic science" entities (kg) and 
"clinical" (umol/L) but surely it is only practical to have them in a single 
schema, whichever it would be?



2) re: the possibility of populating gnumed with a data pack of conversions, 
for which I have shared with Karsten raw data we could use, and of which a 
sample is as follows from which we might take just the units and conversions 
but not the normal ranges:

        Acetaminophen   Serum, plasma   10-30   µg/mL   6.614   66-200  µmol/L
        Acetoacetate    Serum, plasma   <1      mg/dL   97.95   <100    µmol/L
        Acetone Serum, plasma   <1.0    mg/dL   0.172   <0.17   mmol/L
        Acid phosphatase        Serum   <5.5    U/L     16.667  <90     nkat/L
        Activated partial thromboplastin time (APTT)    Whole blood     25-40   
s       1       25-40   s
        …
        Glucose Serum   70-110  mg/dL   0.0555  3.9-6.1 mmol/L 

… is it agreed that this first needs the "figuring out" of whether and how we 
can enable tests residing in

        clin.test_type

to key automatically (rather than having to have been statically linked to 
pk_id) to the appropriate "matching" basis for conversion?

… is it flawed or reasonable to think that this might be achievable by the use 
of unique descriptors for related test_types, for example

        any tests which measure length

                height
                head circumference
                abdominal circumference
                calf girth
                Schober test

        could key to the foundational term

                length

        whereas

                glucose, serum
                glucose, plasma
                glucose, whole blood
                glucose, CSF
                glucose, urinary
                glucose, capillary

        could all key to foundational analyte

                glucose

        and all sharing, in common,

                mg/dL * 0.0555 --> mmol/L
                mmol/L / 0.0555 --> mg/dL

… how could non-English version support be handled in GNUmed?

… does the sample data point to the possible value of supporting to store, 
within clin.test_type, a column / value for the type of specimen, such as

        serum
        urine

??

-- Jim